本文首先给出了一种模拟双退火算法,增加第二退火温度,随着第二温度的降低,逐渐减少状态向量中可变元素的个数,直至只允许一个元素变化,由此可快速精确的逼近最优解的,其次利用该算法对模糊CMAC(小脑模型关节控制器)神经网络的权值进行优化,并构成了一种新型模糊CMAC神经网络。在模型未知的情况下,利用该神经网络对一个非线形时变模型进行控制仿真,效果很好。 In this paper, a simulated annealing algorithm is proposed to increase the second annealing temperature. With the decrease of the second temperature, the number of the variable elements in the state vector is gradually reduced until only one element change is allowed, thereby making it quick and accurate Then the algorithm is used to optimize the weights of the fuzzy CMAC neural network and constitutes a new type of fuzzy CMAC neural network. In the case of unknown model, it is very effective to control the simulation of a nonlinear time-varying model by using the neural network.